Search

Accurate and cost-effective quantification of the carbon cycle for agroecosystems at decision-relevant scales is critical to mitigating climate change and ensuring sustainable food production. However, conventional process-based or data-driven modeling approaches alone have large prediction uncertainties due to the complex biogeochemical processes to model and the lack of observations to constrain many key state and flux variables. Here we propose a Knowledge-Guided Machine Learning (KGML) framework that addresses the above challenges by integrating knowledge embedded in a process-based model, high-resolution remote sensing observations, and machine learning (ML) techniques. Using the U.S. Corn Belt as a testbed, we demonstrate that KGML can outperform conventional process-based and black-box ML models in quantifying carbon cycle dynamics. Our high-resolution approach quantitatively reveals 86% more spatial detail of soil organic carbon changes than conventional coarse-resolution approaches. Moreover, we outline a protocol for improving KGML via various paths, which can be generalized to develop hybrid models to better predict complex earth system dynamics.

As one of the major agricultural production areas in the world, the United States (U.S.) Midwest plays a vital role in the global food supply and agricultural ecosystem services. Although significant efforts have been made in modeling the carbon cycle dynamics over this area, large uncertainty still exists in the previous simulations in terms of reproducing individual components of the carbon cycle and their responses to environmental variability. Here we evaluated the performance of an advanced agroecosystem model, ecosys, in simulating carbon budgets over the U.S. Midwest, considering both the magnitude of carbon flux/yield and its response to environmental (climate and soil) variability. We conducted model simulations and evaluations at 7 cropland eddy-covariance sites as well as over 293 counties of Illinois, Indiana, and Iowa in the U.S. Midwest. The site-level simulations showed that ecosys captured both the magnitude and seasonal patterns of carbon fluxes (i.e., net ecosystem carbon exchange (NEE), ecosystem gross primary production (GPP), and ecosystem respiration (Reco)), leaf area index (LAI), and dynamic plant carbon allocation processes, with R2 equal to 0.92, 0.87, 0.87, and 0.78 for GPP, NEE, Reco, and LAI, respectively across all the sites compared with the observations. For regional scale simulations, ecosys reproduced the spatial distribution and interannual variability of corn and soybean yields with the constraints of observed yields and a new remotely sensed GPP product, with R2 of multi-year averaged simulated and observed yield equal 0.83 and 0.80 for corn and soybean, respectively. The simulated responses of carbon cycle dynamics to environmental variability were consistent with that from the empirical observations at both site and regional scales. Our results demonstrated the applicability of ecosys in simulating the carbon cycle and soil carbon dynamics of the U.S. Midwestern agroecosystems under different climate and soil conditions.

Recurrent extreme landscape fire episodes associated with drought events in Indonesia pose severe environmental, societal and economic threats. The ability to predict severe fire episodes months in advance would enable relevant agencies and communities more effectively initiate fire preventative measures and mitigate fire impacts. While dynamic seasonal climate predictions are increasingly skilful at predicting fire-favourable conditions months in advance in Indonesia, there is little evidence that such information is widely used yet by decision makers. In this study, we move beyond forecasting fire risk based on drought predictions at seasonal timescales, and (i) develop a probabilistic early fire warning system for Indonesia (ProbFire) based on multilayer perceptron model using ECMWF SEAS5 dynamic climate forecasts together with forest cover, peatland extent and active fire datasets that can be operated on a standard computer, (ii) benchmark the performance of this new system for the 2002-2019 period, and (iii) evaluate the potential economic benefit such integrated forecasts for Indonesia. ProbFire's event probability predictions outperformed climatology-only based fire predictions at three to five-month lead times in south Kalimantan, south Sumatra and south Papua. In central Sumatra, an improvement was observed only at one month lead time, while in west Kalimantan seasonal predictions did not offer any additional benefit over climatology only-based predictions. We (i) find that seasonal climate forecasts coupled with the fire probability prediction model confer substantial benefits to a wide range of stakeholders involved in fire management in Indonesia and (ii) provide a blueprint for future operational fire warning systems that integrate climate predictions with non-climate features. [ABSTRACT FROM AUTHOR]

Water table depth (WTD) is the predominant biophysical control over the occurrence of peat and forest fires and greenhouse gas emissions in tropical peatlands. In Indonesia, prolonged droughts caused by El-Niño and/or positive Indian Ocean Dipole, exacerbated by extensive peat drainage for agriculture and plantation establishment, can promote severe peatland fires by lowering WTD and hence desiccating surface and sub-surface peats. These peatland fires create local health hazards and cause irreversible damages to peatland ecosystems. Severe El Niño episode of 2015 led to a major and damaging increase in Indonesian peatland fires, highlighting an urgent need to develop operational systems to forecast potentially severe fire events to mitigate the impacts of fire and haze. The 2002 "ASEAN Agreement on Transboundary Haze Pollution", signed and ratified by a total of 10 ASEAN states, including Indonesia since 2014, identifies a critical need for such systems based on near-time climate projections (e.g. seasonal forecasts). However, such systems have not yet been developed. While an operational early warning system for forecasting dangerous burning conditions is currently within reach using state-of-the-art modelling tools, such as the ECMWF's System 5 seasonal forecast model (SEAS5), development is still hampered by insufficient knowledge about the influence of fluctuations in peat moisture on fire, particularly during periods of extreme drought (e.g. 1997-98 and 2015 El Niño episodes). The main objectives of this study were: i) initialise and run the process-based ecosystem model "ecosys" to study how WTD and peat moisture profiles change in tropical peatlands across Riau province, Sumatra, in response to drought and land cover change, focusing on the 2015 El-Niño event; and ii) examine whether those changes could have been predicted using SEAS5. "Ecosys" has been rigorously tested against high resolution site-level observations over tropical peatlands. Model spin-up from 2008-2014 was driven by inputs from ECMWF's climate reanalysis data (ERA5), followed by a comparison of three model runs for 2015, driven by i) ERA5, ii) ERA5 climatology, and iii) SEAS5 hindcasts. Model inputs for peat soil properties, peat depths and land use and land cover (LULC) details were gathered by using several mosaics from published regional and global soil and LULC datasets. The modelled landscape for Riau province was divided into 5kmx5km grid-cells excluding open water bodies, and seasonally-flooded and urban areas. Model inputs were up- or down-scaled according to grid cell size of the modelled landscape. The model outputs of peat moisture profiles and WTD showed how peat moisture and WTD were significantly affected by weather and land uses during the dry season of 2015, and these results were predictable based on SEAS5. Our work is a pioneering attempt to perform large-scale process-based modelling to predict seasonal variations in tropical peatland WTD. We report on the benchmarking of "ecosys" outputs against remotely sensed datasets e.g. Soil Moisture Active Passive and site-level monitoring networks. This study forms part of the Towards a Fire Early Warning System for Indonesia (ToFEWSI) project (2017-2020), funded by UK's National Environment Research Council (NERC) and Indonesia Endowment for Education (LPDP). [ABSTRACT FROM AUTHOR]